writing-plans
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external project specifications and generated plans, which presents an attack surface where malicious or adversarial instructions embedded in those documents could potentially influence the behavior of the agent during the review or planning phase.
- Ingestion points: Input files are specified via placeholders
[PLAN_FILE_PATH]and[SPEC_FILE_PATH]insideplan-document-reviewer-prompt.md. - Boundary markers: There are no explicit text delimiters or isolation guardrails surrounding the file placeholders to separate the content from the prompt's instructions.
- Capability inventory: The skill writes implementation plan documents to the local file system at
docs/superpowers/plans/paths. - Sanitization: No input validation, escaping, or structural sanitization is defined for the ingested files before they are processed by the reviewer subagent.
Audit Metadata